In the constellation Virgo, the galaxy NGC 4254 drags a tail of cold gas. This means it is speeding into the superheated medium of the cluster and losing its fuel for star formation. But it could have already passed through here once — and then its current appearance is deceptive. Telling a 'newcomer' from a 'returnee' from images is nearly impossible, yet understanding how the environment reshapes galaxies depends on it.

Astronomers trained a computer program on hundreds of simulations where all orbits are known. The program is a 'random forest' (an algorithm that aggregates answers from many simple decision rules and produces a final estimate). It analyzes what can be measured from Earth: distance to the cluster center, radial velocity (determined from the spectrum), direction of motion across the sky, number of nearby neighbors. Even if some data is missing, the forest adapts and still distinguishes returnees from newcomers with about 84% accuracy.

When the program was applied to real galaxies with tails in the Virgo cluster, all turned out to be newcomers. That means the superheated gas strips away cold fuel on the first pass, and the galaxy gets no second chance. Now, by separating clean groups of returnees and newcomers, scientists can understand step by step how clusters quench star formation.